Efficient design optimization of microwave circuits using parallel computational methods

An efficient design optimization technique is proposed for microwave circuits using multi-point function evaluations in each iteration to achieve a large effective optimization update. In this technique, a local model is developed in the neighborhood of the nominal point for each iteration. EM data for building the local model is optimally chosen using design of experiments method. Further, the EM data is generated by running multiple EM simulations in parallel on a multi-processor environment. Hence, we can afford to develop a local model for a wider neighborhood by increasing the number of EM simulations and still keep the computation time low. Once developed, the local model allows a large and effective optimization update in the wider neighborhood of the nominal point, thereby resulting in fewer optimization iterations and achieving speedup for the optimization process. Two typical band-pass filter examples are used to illustrate the proposed technique.